All notable changes to this project will be documented in this file.
The format is based on Keep a Changelog, and this project adheres to Semantic Versioning.
0.1.0 - 2024-12-31
- Initial release of Easy Debug
easy_dump: Dump tensors, arrays, and data structures with automatic format detectioneasy_view: Inspect binary files (.pt, .npy, .json, .pkl) in readable formateasy_compare: Compare files and directories with detailed diff reports- Automatic file extension detection based on data type
- Support for PyTorch tensors, NumPy arrays, JSON, pickle, and text formats
- Rich statistics for tensors and arrays
- Environment variable configuration (
ENABLE_EASY_DUMP,EASY_DUMP_PATH) - .env file support for custom dump paths
- Run ID management for organizing dumps by experiment
- Comprehensive documentation and examples
- Apache 2.0 license
- Smart file naming with training step, function name, keys, and UUID
- Recursive comparison of nested data structures
- Folder comparison mode for comparing entire experiment directories
- Key filtering for JSON file comparisons
- Detailed and summary comparison modes
- CLI tools for viewing and comparing dumps
- Type-safe value formatting with error handling
basic_usage.py: Basic dumping and viewing operationstraining_loop.py: Integration with PyTorch training loopscomparison_workflow.py: Experiment comparison workflow